| 4 |
Notes on tags used in MITgcmUV |
Notes on tags used in MITgcmUV |
| 5 |
============================== |
============================== |
| 6 |
|
|
| 7 |
o adding precip control xx_precip |
o gmredi: fix bugs in slope tapering ldd97 (depthZ had the wrong sign) |
| 8 |
o moved ptracer ctrl index 17 -> 32 |
and update lab_sea output.txt file. |
| 9 |
|
o completing 3-dim. relaxation code: additional term |
| 10 |
|
1/tau*( theta(i,j,k) - thetaStar(i,j,k) ) |
| 11 |
|
in S/R EXTERNAL_FORCING_T (file external_forcing.F) |
| 12 |
|
* At compile time, in "EXF_OPTIONS.h" set |
| 13 |
|
#define ALLOW_CLIMTEMP_RELAXATION |
| 14 |
|
#define ALLOW_CLIMSALT_RELAXATION |
| 15 |
|
* At runtime in "data", under &PARM03 set (e.g., units in seconds) |
| 16 |
|
tauThetaClimRelax3Dim = 15552000.0, |
| 17 |
|
tauSaltClimRelax3Dim = 15552000.0, |
| 18 |
|
* 3 cases: |
| 19 |
|
1) Use a constant-in-time 3-dim. field |
| 20 |
|
2) Provide 12 3-dim. fields corresponding to Jan., ... , Dec. |
| 21 |
|
3) Real-time relax. to time-evolving prescribed field |
| 22 |
|
|
| 23 |
|
checkpoint57y_post |
| 24 |
|
First step for a NLFS adjoint |
| 25 |
|
o initially suppress rStar (new flag DISABLE_RSTAR_CODE) |
| 26 |
|
o new init. routines for calc_r_star, calc_surf_dr |
| 27 |
|
o still need to deal with ini_masks_etc |
| 28 |
|
o testreport seemed happy |
| 29 |
|
checkpoint57y_pre |
| 30 |
|
|
| 31 |
|
o obcs: always reset wVel & eta (to zero or prescribe it) at the OBs. |
| 32 |
|
o add new package rbcs (for relaxing ptracers at boundaries/depth/surface etc) |
| 33 |
|
o add new offset feature to offline pkg (plus minor cleanup) |
| 34 |
|
o cfc fixes for final year of integration |
| 35 |
|
o clean up GCHEM_OPTIONS.h/DIC_OPTIONS.h (all options should |
| 36 |
|
be in GCHEM_OPTIONS). |
| 37 |
|
|
| 38 |
|
checkpoint57x_post |
| 39 |
|
o fix testreport Pb in multiple test experiment. |
| 40 |
|
o cleanup our namelist syntax in verification: |
| 41 |
|
- make it more F95 standards compliant (some work still needed) |
| 42 |
|
- allows the vast majority of our tests to work with gfortran |
| 43 |
|
o Non-Hydrostatic: |
| 44 |
|
- put all NH variables (formerly in DYNVARS.h & GW.h) in NH_VARS.h |
| 45 |
|
- fix exchange calls for CS-grid using the new EXCH_S3D_RL. |
| 46 |
|
o Allow to apply AB on T,S rather than on AB(gT,gS): |
| 47 |
|
- implemented within #ifdef ALLOW_ADAMSBASHFORTH_3 |
| 48 |
|
- use the same arrays (gtNm,gsNm) to hold tracer field at previous |
| 49 |
|
time-steps (if AB(T,S)) and tendencies (if AB(gT,gS)). |
| 50 |
|
- need a "clever" read_pickup to restart from the other AB formulation |
| 51 |
|
o move local commom bloc /GlobalLo/ (used for singleCpuIO) in EESUPPORT.h |
| 52 |
|
o remove some unused variables (reduces number of compiler warnings) |
| 53 |
|
o adjoint |
| 54 |
|
* Disable useVariableViscosity for adjoint for time being |
| 55 |
|
(there are a number of extensive recomputations) |
| 56 |
|
* Initialize certain variables needed for adjoint |
| 57 |
|
|
| 58 |
|
checkpoint57w_post |
| 59 |
|
o pkg/mdsio: use MAX_LEN_FNAM (instead of hard coded 80) for file-names |
| 60 |
|
fix the singleCpuIO RS version of MDSWRITEFIELD |
| 61 |
|
and add a WARNING if using globalFiles=T with multi-processors. |
| 62 |
|
o write grid-file(s) in only one format (MDS or MNC): default=MNC (if |
| 63 |
|
using MNC); can force MDS output by setting writegrid_mnc=F in data.mnc |
| 64 |
|
o tracer advection scheme: |
| 65 |
|
add DST-2 (=Lax-Wendroff) & 1rst order upwind schemes (the way to get |
| 66 |
|
those 2 schemes before, changing GAD_FLUX_LIMITER.h + to recompile, |
| 67 |
|
was not very convenient). |
| 68 |
|
add implicit version for DST-3 & DST3_Flux_Limit vertical advection; |
| 69 |
|
(further tests needed for the implicit DST & Flux-Limiter schemes). |
| 70 |
|
o PTRACERS: added namelist parameter PTRACERS_ref in analogy to tRef |
| 71 |
|
and sRef for convenience, handy, when you want to initialize |
| 72 |
|
ptracers in a large domain with a constant non-zero value. Also, |
| 73 |
|
untangle ptracers_init.F a little. |
| 74 |
|
|
| 75 |
|
checkpoint57v_post |
| 76 |
|
o change dome exp. to use variable viscosity in side-drag terms. |
| 77 |
|
o remove all the files from the cfc_offline test since they are now |
| 78 |
|
incorporated into tutorial_cfc_offline which has been producing |
| 79 |
|
identical results for a few days of testing |
| 80 |
|
o add new parameter: sideDragFactor to enable half-slip-side BC |
| 81 |
|
and change side-drag S/R to use viscosity from mom_calc_visc |
| 82 |
|
(can switch back by setting sideDragFactor=0.); affects results at |
| 83 |
|
truncation level; regenerate output for exp.: global_ocean.cs32x10(x3), |
| 84 |
|
global_ocean_ebm, global_ocean_pressure, matrix_example & MLAdjust(x5). |
| 85 |
|
o vorticity & strain are now computed without masking, to account for |
| 86 |
|
no-slip-side BC in variable viscosity calculation ; mask is applied |
| 87 |
|
anyway later on, before computing advection/coriolis terms. |
| 88 |
|
o modify exp4 to test new obcs support for reading from external files |
| 89 |
|
and for passive tracers |
| 90 |
|
o fix behavior of ptracers_monitor when monitor output is written to a |
| 91 |
|
netcdf-file; still not independent of monitor.F (writes in the same |
| 92 |
|
file). |
| 93 |
|
o change default value of tRef (put a uniform, meaningful value for ocean |
| 94 |
|
& atmos) and initialize sRef (was missing, as Martin mentioned). |
| 95 |
|
o Redefine viscC?Leith*. Change the coefficients of Leith viscosity |
| 96 |
|
to be proportional to enstrophy dissipation length. Thus, |
| 97 |
|
viscC?leith?_new=pi*(viscC?leith?_old)^(1/3) |
| 98 |
|
This is now consistent with Smagorinsky and with the manual. |
| 99 |
|
|
| 100 |
|
checkpoint57u_post |
| 101 |
|
o OBCS and PTRACERS: add open boundary support for passive tracers |
| 102 |
|
- either use homogenous (pseudo) v.Neumann conditions or prescribe |
| 103 |
|
OB-values from file; this is not different from the way theta and salinity |
| 104 |
|
are treated |
| 105 |
|
- however, Orlanski-radiation conditions are not supported, and the model |
| 106 |
|
will stop if you use pTracers and Orlanski at the same time. |
| 107 |
|
- beefed up the rountine obcs_external_fields_load: now only those open |
| 108 |
|
boundary values are overwritten with values from files for which there |
| 109 |
|
are really files, otherwise the OB-fields remain untouched. This makes |
| 110 |
|
it possible to use different OBs at different ends of the domain (as |
| 111 |
|
with EXF) |
| 112 |
|
o reduce number of exch calls in shap_filt (s2). |
| 113 |
|
o CS-grid: fix divergence contribution to Leith viscosity. Affects (@ |
| 114 |
|
truncation level) 4/5 MLAdjust test results (11&12 digits match). |
| 115 |
|
|
| 116 |
|
checkpoint57t_post |
| 117 |
|
o test biharmonic viscosity + no-slip with exp: global_ocean.cs32x15.viscA4 |
| 118 |
|
o finish adding brackets (coriolis S/R, vecinv) in multi-terms sum to get |
| 119 |
|
same truncation, independent of the face number: affects affects several |
| 120 |
|
test. exp.: Update global_ocean.cs32x15 (+.thsice) & MLAdjust (all 5). |
| 121 |
|
o fix bug in mom_v_sidedrag (biharmonic viscosity was counted 2 times); |
| 122 |
|
note that biharmonic viscosity + no-slip was not tested in any experiment. |
| 123 |
|
o fix bug in non-linear free-surf side-drag ; generate new output for |
| 124 |
|
global_ocean.cs32x15(+.thsice) & global_ocean_pressure. |
| 125 |
|
o make mnc honor the writeBinaryPrec flag for most kinds of output |
| 126 |
|
o eesupp: split EEPARAMS into EESIZE/EEPARAMS and add a vector global sum |
| 127 |
|
o minor change in averaging W (mom_vi_?_vertshear.F) that affects truncation |
| 128 |
|
error; update output from global_ocean.cs32x15(+.thsice) & MLAdjust (all 5). |
| 129 |
|
o add brackets in multi-terms sum to get same truncation, independent of |
| 130 |
|
the face number (manage to get exactly same value of gU,gV at the cs-edge) |
| 131 |
|
affects several test. exp. Update global_ocean.cs32x15 (+.thsice). |
| 132 |
|
o Clean, new version of mom_calc_visc. Adds a new feature, viscAhRemax |
| 133 |
|
and viscA4Remax, which allow one to limit the minimum viscosity by specifying |
| 134 |
|
a maximum grid-scale Reynolds number. Also, changes all length calculations |
| 135 |
|
in the case of grid anisotropy to geometric mean of dx**2 and dy**2. |
| 136 |
|
Suggested values of viscosity parameters found in mom_calc_visc. |
| 137 |
|
Many new viscosity diagnostics available. |
| 138 |
|
|
| 139 |
|
checkpoint57s_post |
| 140 |
|
o Add clearing-house for viscosity calculations, mom_calc_visc and |
| 141 |
|
correct numerous bugs in LeithD and Smagorinsky discretizations. |
| 142 |
|
mom_calc_visc now allows for Smagorinsky, Leith and LeithD to be |
| 143 |
|
used simultaneously, and to be used in mom_vecinv and mom_fluxform. |
| 144 |
|
Mom_vecinv can be used with standard viscous calculation (mom_vi_hdissip.F) |
| 145 |
|
and with Strain-Tension viscous term calculation (mom_hdissip.F). |
| 146 |
|
The Strain-Tension form is now called with useStrainTensonVisc=.true. |
| 147 |
|
ViscAhStrain and viscAhTension are now retired. |
| 148 |
|
o diagnostics "VISCAHD" & "VISCA4D" replace "VISCAH" & "VISCA4". |
| 149 |
|
o In response to JMC's bug report, genmake2 now ignores any soft-links |
| 150 |
|
that it finds in the *current* directory when searching for files. |
| 151 |
|
Soft-links found in any other source directories will still work as |
| 152 |
|
they did before. |
| 153 |
|
o add HAVE_SETRLSTK define and useSETRLSTK flag (eedata) which calls |
| 154 |
|
a C routine to unlimit the stack size |
| 155 |
|
o mnc-ify pkg/aim snapshot vars |
| 156 |
|
o mnc-ify pkg/land as requested by Daniel |
| 157 |
|
o various changes to mnc including: |
| 158 |
|
- all files use the new "BASENAME[[.ITER].{t|f}NUM].nc" format |
| 159 |
|
- output can now be grouped so that all files within a group |
| 160 |
|
change the ITER portion of their names in lock-step together |
| 161 |
|
- can now read ("global") PER-FACE (in addition to PER-TILE) files |
| 162 |
|
and works with both EXCH1 and EXCH2 (but needs more testing) |
| 163 |
|
- writing works for all verification test cases w/ g77 on Linux |
| 164 |
|
|
| 165 |
|
checkpoint57r_post |
| 166 |
|
o make MNC variable initialization happen earlier |
| 167 |
|
o add diagnostics for (almost) each term of the momentum Eq. |
| 168 |
|
o remove SST,SSS,UVEL_k2,VVEL_k2 diagnostics. |
| 169 |
|
o move pkg timeave out of pkg-groups "oceanic" & "atmospheric" (pkg/pkg_groups) |
| 170 |
|
o remove MINIMAL_TAVE_OUTPUT option ; |
| 171 |
|
add "dumpInitAndLast" to replace "nodumps" (and MINIMAL_TAVE_OUTPUT). |
| 172 |
|
o move some basic S/R (used elsewhere in the code) from pkg/mdsio to eesupp/src |
| 173 |
|
allows to compile without pkg/mdsio (& test successfully hs94.1x64x5) |
| 174 |
|
o add MNC output for adjoint of EXF (adexf), SEAICE (adseaice) |
| 175 |
|
o add time-dependent SST, SSS control |
| 176 |
|
o cost terms in ecco/ modularized (new cost_generic.F) |
| 177 |
|
o cost_averages routines are modularized (new cost_averagesgeneric.F) |
| 178 |
|
o new infrastructure for seaice model vs. obs misfits in seaice/ |
| 179 |
|
* driver from the_main_loop.F is seaice_cost_driver.F |
| 180 |
|
* first term is AREA vs. SMR observation misfit (ALLOW_SEAICE_COST_SMR_AREA) |
| 181 |
|
* gradient check in Exp. 6a of natl_box_adjoint/ |
| 182 |
|
o natl_box_adjoint input_seaice for Lab. Sea now available online via wget |
| 183 |
|
(see README file) |
| 184 |
|
|
| 185 |
|
checkpoint57q_post |
| 186 |
|
o fix initialisation Pb in calc_gw (get NANs in exp5 with g77) |
| 187 |
|
o fix annoying months-old bug in mnc/mdsio output for seaice pkg |
| 188 |
|
- bug was reported by Jean-Michel and Ian Fenty |
| 189 |
|
o fix initialization Pb in AIM-Ocean coupled model (aim_sice2aim.F) |
| 190 |
|
update exp. cpl_aim+ocn output. |
| 191 |
|
o Implement 4-level checkpointing: AUTODIFF_4_LEVEL_CHECKPOINT |
| 192 |
|
Needed for high-res. runs with large global files |
| 193 |
|
to be able to limit tapelev size |
| 194 |
|
o Changed some tests for adjoint: |
| 195 |
|
* carbon/ now tests AUTODIFF_2_LEVEL_CHECKPOINT |
| 196 |
|
* natl_box_adjoint/code_bulk/ tests AUTODIFF_4_LEVEL_CHECKPOINT |
| 197 |
|
* global_ocean.90x40x15/ back to default 3-level checkpointing |
| 198 |
|
o Modifications to enable scripted COMMON to MODULE conversion |
| 199 |
|
* verification/OpenAD/code/ has necessary scripts and description; |
| 200 |
|
* converts, compiles and runs successfully; |
| 201 |
|
|
| 202 |
|
checkpoint57p_post |
| 203 |
|
o reading of delXfile & delYfile moved from ini_parms.F to ini_grid.F |
| 204 |
|
o move ${PKG}_MNC_INIT from ${PKG}_READ_PARAMS to ${PKG}_INIT_FIXED |
| 205 |
|
(already the case for some pkgs, including recent MNC_init, e.g. thsice) |
| 206 |
|
o call CALC_GW from DYNAMICS (instead of from FORWARD_STEP) |
| 207 |
|
o Dissipation & phiHyd gradient are always added to gU,gV in timestep.F |
| 208 |
|
(was already the case for dissipation with mom_vecinv, |
| 209 |
|
and also the case for grad.PhiHyd when staggered-timeStep) |
| 210 |
|
This will allow to put dissipation out-off the AB time-stepping. |
| 211 |
|
Unfortunately, affects truncation error. Update output of exp. that "fails" |
| 212 |
|
(cg2d matching number) and leave the others for later: front_relax(10), |
| 213 |
|
global_ocean_pressure(10),ideal_2D_oce(10),internal_wave(9),lab_sea(12), |
| 214 |
|
matrix_example(7!),natl_box(12),plume_on_slope(11),vermix(12x3,11x2). |
| 215 |
|
|
| 216 |
|
checkpoint57o_post |
| 217 |
|
o fix MNC/MDSIO flags in PTRACERS so that (1) it works, and |
| 218 |
|
(2) its ready to be documented during the 20050801 DocFest |
| 219 |
|
o MNC-ify the gmredi package -- tested and works w/ and w/o mnc |
| 220 |
|
for ideal_2D_oce and lab_sea on linux_ia32_g77 |
| 221 |
|
o set to #undef CTRL_SET_OLD_MAXCVARS_30 |
| 222 |
|
in global1x1_tot/code_seaice_adj to access indices 32, 33 |
| 223 |
|
(worth another tag) |
| 224 |
|
|
| 225 |
|
checkpoint57n_post |
| 226 |
|
o adding precip control xx_precip 32 |
| 227 |
|
o adding swflux control xx_swflux 33 |
| 228 |
|
o Replacing 10 forcing penalty cost routines by 1 generic. |
| 229 |
|
o Implementing partitioning in time-mean and variable adjustment |
| 230 |
|
for atmos. state. |
| 231 |
|
o Time-mean constant error is now read in data.ecco (instead of hard-coded) |
| 232 |
o in natl_box_adjoint: |
o in natl_box_adjoint: |
| 233 |
* updated: results_bulk/ |
* updated exp 1 to 10 |
| 234 |
* not yet updated: results/, results_seaice/ |
(but does not have time-mean vs. variable split results yet) |
| 235 |
|
o updated global1x1/*_seaice_adj/ |
| 236 |
|
to incorporate latest changes |
| 237 |
|
o enable useSingleCpuIO for mdsio vector routines |
| 238 |
|
(applies in particular to tapelev I/O) |
| 239 |
|
o removed inefficient K-loop in all active_file*control*.F |
| 240 |
|
|
| 241 |
checkpoint57m_post |
checkpoint57m_post |
| 242 |
o re-arrange exp. adjustment.cs-32x32x1 (adjust_nlfs.cs-32x32x1 merge into) |
o re-arrange exp. adjustment.cs-32x32x1 (adjust_nlfs.cs-32x32x1 merge into) |
Parent Directory
| 
Revision Log
| 
Revision Graph
| 
Patch